Chile stands out for its renewable energy resources and its commitment to developing green hydrogen. However, achieving cost parity with gray hydrogen remains an obstacle, mainly due to high capital costs and sensitivity to scale. This study assesses the technical and economic feasibility of green hydrogen production, using five different plants located in the Magallanes region in the south of the country as a reference. The model integrates a detailed framework of wind generation, PEM electrolysis, compression, and high-pressure storage subsystems, as well as a stochastic economic layer that combines the CAPEX, NPV, and LCOH assessments using Monte Carlo simulations. It also incorporates real-world capacity distributions and probabilistic fluctuations in systems. A sensitivity analysis confirms production scale as the main factor affecting profitability, with a break-even threshold of 0.5 MW. The results show that the LCOH decreases from 7.1 USD to 3.4 USD/kgH2 as capacity increases. The analysis reveals that only 23.88% of small-scale configurations yield positive NPV, underscoring the need for scaling to achieve economic viability.
Accurate 3D semantic occupancy perception is essential for autonomous driving in complex environments with diverse and irregular objects. While vision-centric methods suffer from geometric inaccuracies, LiDAR-based approaches often lack rich semantic information. To address these limitations, MS-Occ, a novel multi-stage LiDAR-camera fusion framework which includes middle-stage fusion and late-stage fusion, is proposed, integrating LiDAR's geometric fidelity with camera-based semantic richness via hierarchical cross-modal fusion. The framework introduces innovations at two critical stages: (1) In the middle-stage feature fusion, the Gaussian-Geo module leverages Gaussian kernel rendering on sparse LiDAR depth maps to enhance 2D image features with dense geometric priors, and the Semantic-Aware module enriches LiDAR voxels with semantic context via deformable cross-attention; (2) In the late-stage voxel fusion, the Adaptive Fusion (AF) module dynamically balances voxel features across modalities, while the High Classification Confidence Voxel Fusion (HCCVF) module resolves semantic inconsistencies using self-attention-based refinement. Experiments on two large-scale benchmarks demonstrate state-of-the-art performance. On nuScenes-OpenOccupancy, MS-Occ achieves an Intersection over Union (IoU) of 32.1% and a mean IoU (mIoU) of 25.3%, surpassing the state-of-the-art by +0.7% IoU and +2.4% mIoU. Furthermore, on the SemanticKITTI benchmark, our method achieves a new state-of-the-art mIoU of 24.08%, robustly validating its generalization capabilities.Ablation studies further confirm the effectiveness of each individual module, highlighting substantial improvements in the perception of small objects and reinforcing the practical value of MS-Occ for safety-critical autonomous driving scenarios.
Tactile sensing is a fundamental modality for embodied intelligence, offering unique and direct feedback on contact geometry, material properties, and interaction dynamics that remote sensors cannot replace. However, unimodal tactile perception is inherently limited by its sparse spatial coverage and lack of global semantic context. With the recent explosion in deep learning and large language models, integrating tactile with vision and language has become essential to bridge physical interaction with semantic reasoning, leading to the emergence of Multimodal Tactile Fusion. Despite rapid progress, the existing researches remain fragmented across disparate datasets, sensing modalities, and tasks, lacking a unified theoretical framework. To address this gap, this paper provides a comprehensive survey of multimodal tactile fusion research up to the first quarter of 2026. We propose a hierarchical taxonomy that organizes the field into two primary dimensions: multimodal datasets and multimodal methods. On the data side, we categorize resources ranging from Tactile-Vision datasets, Tactile-Language datasets, Tactile-Vision-Language datasets, and Tactile-Vision-Other datasets. On the method side, we structure prior work into three core pillars: (1) Multimodal Perception and Recognition, which focuses on object understanding and grasp prediction; (2) Cross-Modal Generation, focusing on bidirectional translation between tactile, vision, and text; and (3) Multimodal Interaction, emphasizing feedback control and language-guided manipulation. Furthermore, we summarize representative tactile sensing hardware, review commonly used evaluation metrics and benchmark settings, and discuss current challenges and promising future directions.
Manual grading of dental students’ embroidery assignments is not only labor-intensive but also subjective. To address these limitations, our study proposes an automated grading model based on ResNet-50 architecture enhanced with a multi-region aggregation mechanism. This approach aims to standardize the grading process, improve fairness and efficiency in assessment. A total of 381 embroidery assignment images were collected from the 2020–2023 student cohorts. The 2022 cohort was designated as an external test set to assess model generalization with different data distributions. We proposed a multi-region aggregation mechanism based on ResNet-50 and compared two aggregation strategies: multi-head attention (MHA) aggregation and average weighting (AW) aggregation. VGG-16, DenseNet-121, ViT, and ResNet-50 were considered as baseline models. All models were trained using 5-fold cross-validation, employing a weighted CrossEntropyLoss to address class imbalance, with evaluation metrics including accuracy, precision, recall, and F1 score. The ResNet-50 AW model achieved the highest test accuracy of 80% on the test set, while the ViT and the VGG-16 models achieved 75%, second to ResNet-50 AW. Although models’ performance degraded on the external test set, ResNet-50 AW maintained the highest accuracy of 64% and reduced misclassifications of grade B and C samples. Despite excelling on the validation set, ResNet-50 MHA showed similar performance to ResNet-50 on the test set. ViT and VGG-16 achieved higher accuracy for grade A on both the test set and the external test set. The ResNet-50 AW model highlights the potential of deep learning methods to automate the grading of artistic assignments via a multi-region aggregation mechanism. Further validation of the model’s generalization is needed. Future work should improve dataset quality and diversity and enhance system interpretability to refine the grading process for greater accuracy and transparency.
Read moreThis paper is concerned with the problem of designing distributed event-triggered H∞ filters over sensor networks subject to heterogeneous coupling intercommunication delays. A new distributed event-triggered scheme is proposed to determine whether or not each sensor's current sampled data should be broadcasted and transmitted to its underlying neighboring nodes through the communication network. In this scheme, each sensor node is able to make its own decisions to broadcast and transmit only when its local measurement output error exceeds a designed threshold. Heterogeneous coupling delays are incorporated in the intercommunication between the specific sensor node and its interacting neighbors. A refined technique is proposed to realize the complicated decoupling among the exchanged measurement outputs in the presence of coupling intercommunication delays. Then the resulting filter error system is modeled by a new delay system subject to finite time-varying 'state' delays. Based on the Lyapunov-Krasovskii functional method, a sufficient condition for distributed event-triggered H∞ filter design is established, from which the desired filter parameters and the triggering parameter in the event condition can be co-designed. The filter design problem is posed in terms of linear matrix inequalities. A quarter-car suspension model is finally presented to show the effectiveness and feasibility of the developed theoretical results.
Read moreThis paper explores the use of a Deep Reinforcement Learning (DRL) model for dynamic portfolio management in the financial market. With the help of deep neural networks and the Twin-Delayed Deep Deterministic Policy Gradient (TD3) algorithm, the framework is able to process high dimensional market data and dynamic environment. The TD3 algorithm incorporates transaction costs and risk aversion constraints in order to simulate the environment of real-world investments. It uses features such as the Moving Average ConvergenceDivergence (MACD) and Relative Strength Index (RSI) to construct its decision-making state space. The performance of the model was assessed using the historical data of six NASDAQ stocks, starting from 2021 to 2023. The results obtained were then compared with two other methods, namely Proximal Policy Optimization (PPO) and Deep Deterministic Policy Gradient (DDPG). Some of the performance measures of the portfolio include cumulative return, annual volatility, Sharpe ratio and the maximum drawdown. The TD3 algorithm produced better results in terms of cumulative and risk return, where it got a $51.28 \%$ cumulative return as compared to a cumulative return of ${2 5. 9 1 \%}$ by DDPG and $17.56 \%$ by PPO. However, the TD3 managed portfolio was accompanied by high annual volatility and drawdown suggesting a risk-return paradox. From the results, it is shown that the TD3 policy is able to produce high returns while maintaining a certain level of risk and outperform static strategies like buy and hold. However, it also included some drawbacks, where the model was not able to forecast the short-term movements of the market and was based on lagging indicators.
Read moreThis study assesses the technical, operational, environmental, and economic feasibility of integrating alkaline water electrolysis (AEL) using on-site measured surplus electricity from two 20 MW natural-gas turbogenerators installed at a Central Processing Facility (CPF) in a Colombian oilfield. Unlike approaches based on modeled profiles, the analysis relies on more than 31,000 experimental records of gas consumption and active power, enabling an accurate characterization of the structural availability of energy surpluses under real operating conditions. A specialized industrial water treatment and purification company was consulted and provided with the physicochemical characterization results obtained from process water samples analyzed by an accredited laboratory. Based on these parameters, the technical supplier confirmed the feasibility of designing a multistage treatment train, including equalization, filtration, clarification, activated carbon, ultrafiltration, and reverse osmosis, capable of achieving final conductivities at or below 5 µS/cm. This water quality level is compatible with typical industrial alkaline electrolysis requirements and in line with technical specifications commonly aligned with ASTM and ISO standards for pressurized AEL systems. A strategic comparison between PEM and AEL technologies, supported by IFE/EFE matrices and sensitivity analyses, identified alkaline electrolysis as the optimal alternative under a stable electrical profile and capital expenditure constraints. Energy sizing for scenarios between 1.5 and 10 MW, assuming continuous 24 h operation and an average specific consumption of 50 kWh/kg H2, yields productions between 0.5 and 3.5 t H2/day, with electrical efficiencies above 70%. A 20-year financial analysis indicates a techno-economic threshold near 3 MW (NPV > 0; IRR > WACC), with optimal performance in the 6.5–10 MW range and payback periods between 2 and 4 years under internal valorization of the surplus electricity. From an environmental perspective, the produced hydrogen is classified as low-carbon rather than “green” due to its thermal origin; however, the integration improves the turbines’ operating regime and valorizes surplus electrical exergy that was previously unused, providing a replicable strategy for industrial assets with self-generation and treatable water availability.
Read moreThe Hybrid Smart Energy Community (HySEC) model is an integrated framework for optimizing hybrid renewable energy systems, unifying BIM, IoT, and data-driven modeling, as an innovative approach for the energy transition. A Revit—Twinmotion BIM model, enriched with topographic, CAD, and real-image data, enhances spatial accuracy and stakeholder communication, while a digital–physical architecture linking sensors, gateways, edge devices, and cloud platforms enables decentralized peer-to-peer communication and real-time monitoring. The framework is applied to a smart energy community composed of a hydropower–wind–solar PV system serving six buildings (48.8 MWh/year), supported by high-resolution hourly Open-Meteo data. A NARX neural network trained on 8760 hourly observations achieves an MSE of 2.346 at epoch 16, providing advanced predictive capability. Benchmarking against HOMER demonstrates clear advantages in grid exports (15,130 vs. 8274 kWh/year), battery cycling (445 vs. 9181 kWh/year), LCOE (€0.09 vs. €0.180/kWh), IRR (9% vs. 6%), payback (8.7 vs. 10.5 years), and CO2 emissions (−9.4 vs. 101 tons). These results confirm HySEC as a conceptually flexible solution that strengthens energy autonomy, supports heritage site rehabilitation, and promotes sustainable rural development.
Read moreDue to the limited understanding of Industrial Control Systems (ICSs), device identification has become increasingly vital for threat detection and security defense in ICS environments. However, the narrow range of device types and models in the existing datasets has significantly hindered the effectiveness and scalability of current device identification methods. To address this gap, we introduce a novel data collection framework specifically designed for ICS devices and present the resulting dataset, ICSLibrary, which we have made publicly available. To the best of the authors' knowledge, ICSLibrary is the first dataset dedicated to device identification in ICS security. It encompasses the most extensive range of device types, models and instances from 27 industrial vendors, collected across two countries over a 21-month period. Furthermore, we use ICSLibrary as a benchmark to evaluate several typical device fingerprinting methods, revealing a notable 16% drop in accuracy in the device model identification task, which highlights the unique challenges posed by ICSLibrary.
Read moreOnline Multi-Object Tracking (MOT) plays a pivotal role in autonomous systems. The state-of-the-art approaches usually employ a tracking-by-detection method, and data association plays a critical role. This paper proposes a learning and graph-optimized (LEGO) modular tracker to improve data association performance in the existing literature. The proposed LEGO tracker integrates graph optimization, which efficiently formulates the association score map, facilitating the accurate and efficient matching of objects across time frames. To further enhance the state update process, the Kalman filter is added to ensure consistent tracking by incorporating temporal coherence in the object states to further enhance the state update process. Our proposed method, utilising LiDAR alone, has shown exceptional performance compared to other online tracking approaches, including LiDAR-based and LiDAR-camera fusion-based methods. LEGO ranked 3<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><i>rd</i></sup> among all trackers (both online and offline) and 2<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><i>nd</i></sup> among all online trackers in the KITTI MOT benchmark for cars<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup>, at the time of submitting results to KITTI object tracking evaluation ranking board. Moreover, our method also achieves competitive performance on the Waymo open dataset benchmark.
Read moreDraining operations using pressurised air can produce sub-atmospheric pressures that pose a significant risk to structural integrity, given the pipe stiffness class. This research presents a modelling strategy for predicting water velocities during the occurrence of this phenomenon. The proposed approach combines a physically based hydraulic formulation with machine learning techniques for making this prediction. A calibrated rigid water column model is first employed to reproduce the transient interaction between the expanding air phase and the draining water column. Input parameters include pipe bridge height varying from 0.5 to 3.0 m, a valve loss dimensionless coefficient ranging from 2.0 to 14.0, and an initial water column length between 163.0 and 286.3 m. Subsequently, a Monte Carlo scheme is used to generate a representative dataset. A total of 28 models were assessed, among which a wide neural network demonstrated superior predictive capability, achieving root-mean-square error values between 0.043 and 0.056 m/s and coefficients of determination ranging from 0.996 to 0.997 for the validation and testing stages, respectively. Sensitivity analyses indicate that the minor loss coefficient governs the water velocity response, whereas geometric features such as the pipe bridge height exert a comparatively minor influence.
Read more
.jpg){kind=small}